The Study of an Optimal Robust Design and Adjustable Ordering Strategies in the HSCM

Table 4

The neural networks for the model.

Structures (input nodes-hidden nodes-output nodes)	RMSE
	Training	Testing
4-6-1	0.03136	0.02888
4-5-1	0.02872	0.02564
4-4-1	0.02521	0.02338
4-3-1	0.02766	0.02462
4-2-1	0.02899	0.02317
4-1-1	0.03100	0.02891

Note. The learning rate was set to autoadjust to value between 0.01 and 0.3; the momentum coefficient is 0.80; and the number of iterations is 15,000.